Research On Cancelable Finger Vein Template Protection Algorithm

With the widespread application of finger vein recognition systems,their security issues are increasingly apparent.Such as,when the user's finger vein feature is damaged or stolen,the user cannot revoke,change or republish the registered finger vein,and when the finger vein registered in one application is stolen,the finger vein saved in other applications is no longer safe.Therefore,in response to biometric security issues,generation techniques of biometric key,encrypted biometrics,and biometric conversion have been proposed.Among them,the cancelable biometric template,because it does not fully meet the conditions of irreversibility,nonlinkability,revocability and performance preservation and then become the focus of biometric template protection.This article conducts research around cancelable finger vein templates,includes three steps: feature extraction,generate a cancelable template,template matching.The main research contents are summarized as follows:(1)The current cancelable biometric template algorithm with better recognition performance cannot guarantee security,but the algorithm that guarantees the security of biometric templates have poor recognition performance.In order to improve the performance and security of the finger vein feature,a cancelable template for finger vein was proposed.Firstly,extract the feature of the finger vein,find the edge areas in the image and optimize the gradient amplitude according to the position of the pixels in the image to increase discrimination,and differential excitation is the ratio of the improved gradient amplitude to the current pixel;the first and second-largest Gabor orientations are combined with differential excitation to form joint distribution features as the two feature vectors of the finger vein;secondly,reduce the dimensionality of the extracted feature vector by controlling the variance,and combine the reduced feature vector with the pseudo-random matrix generated by token to generate a cancelable template to ensure feature security;finally,based on the improved canonical correlation analysis,the cancelable templates in two orientations are fused to increase the irreversibility.(2)Ranking-based local sensitive hashing algorithm,"Index of Max"(IoM)hashing algorithm,for finger vein template protection.With externally generated random parameters,IoM hashing transforms a real-valued biometric feature vector into discrete index hashed code.Firstly,we show the implementation of the feature extraction of finger vein and apply it to IoM hashing,that is,Gaussian random projection-based and uniformly random permutation-based hashing schemes,which proves the effectiveness of our finger vein feature extraction for IoM;secondly,two improvements were made for IoM hashing: 1)for the extracted feature vector of finger vein,merging the feature and the replacement feature generated by the random permutation set into a feature matrix can further enhance the irreversibility;2)in order to adjust the error generated by the random matrix better and cause the max index to be inaccurate,the largest hash index is extended to the largest and second largest hash index to generate a cancelable finger vein template.The experimental results show there is well accuracy on PolyU and SDUMLA-FV databases.The analyses justify its resilience to the existing and newly introduced security and privacy attacks as well as satisfy the revocability and unlinkability of cancelable biometrics.